High-resolution daily gridded minimum, mean and maximum temperature datasets for Germany and river catchments Stefan Krähenmann, Andreas Walter and Simona Höpp Deutscher Wetterdienst Zentrales Klimabüro 08.09.2017 EMS Annual Meeting
Overview Motivation and concept Data basis Analysis of 2 m temperature Results Evaluation of temperature analysis 08.09.2017 EMS Annual Meeting
Dataset required to bias-correct regional climate projection data Both bias-corrected RCM data and observational grids serve as input data for impact modelling and related studies within Network of Experts The ultimate goal being the determination and implementation of adaptation measures in order to enhance the resilience of roads High-resolution grids of minimum and maximum temperature for various applications including technical meteorology, agricultural meteorology Presentation by S. Hänsel: “Assessing weather related risks to the German transport infrastructure” (EMS2017-606) on Tuesday, 05.09.17 at 15:00–15:15 within session OSA2.2 Spatial Climatology 08.09.2017 EMS Annual Meeting
Requirements on the temperature dataset High spatial resolution and precision True histogram and spatial variability All available observations fulfilling basic quality requirements used Temporal variations in the error structure Not applicable to trend analysis Method for temperature interpolation in complex topography Frei, 2014: Daily temperature maps for Switzerland; non-linear temperature profile & non-Euclidean distances Hiebl, and Frei, 2015: Daily minimum and maximum temperature maps for Austria; non-linear temperature profile, non-Euclidean distances & urban heat island effect Krähenmann et al., 2016: Hourly temperature maps for Germany; non-linear temperature profile, non-Euclidean distance, urban heat island effect and cloudiness 08.09.2017 EMS Annual Meeting
Overview Motivation and concept Data basis Analysis of 2 m temperature Results Evaluation of temperature analysis 08.09.2017 EMS Annual Meeting
08.09.2017 EMS Annual Meeting
Overview Motivation and concept Data basis Analysis of 2 m temperature Results Evaluation of temperature analysis 08.09.2017 EMS Annual Meeting
Configuration and adaptations Daily temperature extremes Sub regions of the background field Urban heat island Accounting for coastal distance Simplification of distance measure 08.09.2017 EMS Annual Meeting
Lapse rates derived in 13 regions Background field Consistency To ensure consistency between TAS, TMIN and TMAX, TAS is primarily interpolated in two steps Lapse rates derived in 13 regions Background field Interpolation of residuals Two-step interpolation of TMIN- and TMAX-deviations to TAS 08.09.2017 EMS Annual Meeting
Non-linear temperature lapse rate Representative on macro scale Background field Non-linear temperature lapse rate Representative on macro scale Temperature inversions of various depths allowed *Frei, 2014 08.09.2017 EMS Annual Meeting
Urban heat island (UHI) UHI effect spoils urban time series Distances between grid node and station increased basing on UHI-potential, depending on City size and location within an agglomeration Time of day and season (alters UHI-based distance penalty) Current weather conditions (alters UHI-based distance penalty) – Berlin – Potsdam – UHI-effect Temperature [°C] Hour since the begin of year 08.09.2017 EMS Annual Meeting
CORINE land use dataset* Maximum UHI effect [°C] Derive maximum UHI effect from CORINE land use 08.09.2017 EMS Annual Meeting *COPERINICUS
Euclidean distance (λ = 0) Cross the mountain Target Down to the valley 08.09.2017 EMS Annual Meeting
Non-Euclidean distance (λ >> 0) Here: Only account for elevation difference Target Go around the mountain Stay on elevation line 08.09.2017 EMS Annual Meeting
Residual field Peak z.pen=150 Coast c.pen=2 Valley z.pen=150 Models local structures not captured by the lapse-rate Cold pools, Foehn effect Weighting scheme derived from modified distances: Distance ~ Δlat, Δlon, z.pen*Δheight, c.pen*Δdist_to_coast, uhi.pen*ΔUHImax z.pen: relative weighting of vertical distance c.pen: relative weighting of distance to coastline uhi.pen: relative weighting of difference in UHImax z.pen/k.pen/uhi.pen = 0 => Euclidian distance Valley z.pen=150 Berlin uhi.pen=0.25 08.09.2017 EMS Annual Meeting
Derivation of distance penalties Independently chosen in four regions Unequal complexity of terrain & proximity to the coast requires definition of sub regions penalty depends on weather conditions penalty depends on difference between air and sea temperature 08.09.2017 EMS Annual Meeting
Overview Motivation and concept Data basis Analysis of 2 m temperature Results Evaluation of temperature analysis 08.09.2017 EMS Annual Meeting
Profiles for daily mean temperature (2.7.1995) UHI- and cold pool - station 08.09.2017 EMS Annual Meeting
Residuals & analysis 2.7.1995 TAS Results Residuals & analysis 2.7.1995 TAS Temperature analysis [°C] 08.09.2017 EMS Annual Meeting
Overview Motivation and concept Data basis Analysis of 2 m temperature Results Evaluation of temperature analysis 08.09.2017 EMS Annual Meeting
Leave-one-out cross validation (1951-2015) Bias TAS (DJF) Bias TAS (JJA) [°C] BIAS MAE TAS -0.01 0.57 TMIN -0.02 0.66 TMAX 0.02 0.61 08.09.2017 EMS Annual Meeting
Evaluation MAE-Improvement [%] due to modified distances vs. eucl. distances (1995-2012) MAE-Improvement [%] when accounting for UHI (1995 – 2012) 08.09.2017 EMS Annual Meeting
Conclusions Temperature dataset Daily mean and extremes Spatial resolution 1 km² and 25 km² 1.Jan.1951 – 31.Dec.2015 Interpolation method suitable for regions with variable topography Methodological adaptations: UHI-effect, distance to coast UHI-correction improved temperature estimates in urban areas MAE 0.6°C for mean temperature and 0.7°C for extremes Consistency of mean temperature and extremes 08.09.2017 EMS Annual Meeting
Literature http://www.bmvi-expertennetzwerk.de/DE/Home/home_node.html Frei, 2014: Interpolation of temperature in a mountainous region using non-linear profiles and non-Euclidian distances. Int. J. Climatol. 24, 1585-1505. Hiebl, and Frei, 2015: Daily temperature grids for Austria since 1961 – concept, creation and applicability. Theor. Appl. Climatol. Doi: 10.1007/s00704-015-1411-4. Krähenmann et al., 2016: High-resolution grids of hourly meteorological variables for Germany. Theor. Appl. Climatol. Doi:10.1007/s00704-016-2003-7. Wienert, U., F. Kreienkamp, A. Spekat, and W. Enke, 2013: A simple method to estimate the urban heat island intensity in data sets used for the simulation of the thermal behavior of buildings. Met. Z., 22(2), 179-185. 08.09.2017 EMS Annual Meeting
Appendix
Evaluation (1995-2015) Bias TMIN (DJF) Bias TMIN (JJA) 08.09.2017 EMS Annual Meeting
Evaluation (1995-2015) Bias TMAX (DJF) Bias TMAX (JJA) 08.09.2017 EMS Annual Meeting
Evaluation (1995-2015) MAE TAS (DJF) MAE TAS (JJA) 08.09.2017 EMS Annual Meeting
Evaluation (1995-2015) MAE TMIN (DJF) MAE TMIN (JJA) 08.09.2017 EMS Annual Meeting
Evaluation (1995-2015) MAE TMAX (DJF) MAE TMAX (JJA) 08.09.2017 EMS Annual Meeting
Annual cycle of distance penalties elevation Dist. to coast UHI coast north south alpine 08.09.2017 EMS Annual Meeting
Overlay of the urban heat island effect (I) Estimation of maximal UHI effect basing on city population Population UHImax [K] 10.000 4,0 100.000 6,0 500.000 7,4 1.000.000 8,0 3.000.000 9,0 UHImax: maximal UHI effect [K] pop: City population UHImax occurs in low-wind, cloudless weather conditions. 08.09.2017 EMS Annual Meeting
Overlay of the urban heat island effect (II) The current UHI depends on the weather condition (wind speed, cloudiness). UHI(t): UHI [K] at t t: time (hour) cf(v24): correction factor due to wind speed (average over 24 h before t) cf(N24): correction factor due to cloudiness (average over 24 h before t ) cf(t): correction factor of UHImax due to its daily cycle 08.09.2017 EMS Annual Meeting
Correction factor cf(t) for June Overlay of the urban heat island effect (III) Approximation of the UHI cycle bases on published observations cf(t) Quelle: CEC Diurnal UHI cycle as observed for Düsseldorf (KUTTLER, 1997) Correction factor cf(t) for June Development of the correction factor cf(t) for daily cycle using a Fourier series: 08.09.2017 EMS Annual Meeting
Overlay of the urban heat island effect (IV) Estimation of the weather influence (cloudiness, wind speed) on UHI Correction factor (cf) depending on wind (24 h average) and city population (Ew) Correction factor (cf) depending on cloudiness (24 h average) cf(v24) cf(N24) Quelle: CEC 08.09.2017 EMS Annual Meeting
Overlay of the urban heat island effect (V) The UHI is modified by the local building structure. 0,5 0,7 1 08.09.2017 EMS Annual Meeting
Estimation of the UHI effect at the stations Assignment of an UHI potential (0-1) depends on land use data CORINE land use data of 100 m spatial resolution (50 classes) 08.09.2017 EMS Annual Meeting
CORINE land use classes for Aachen 1 km 08.09.2017 EMS Annual Meeting
Assignment of UHI potential to CORINE land use classes: Aachen Assignment to land use classes: Continuous urban fabric: 1, discontinuous urban fabric: 0.5, industrial areas: 0.7, green urban areas: 0.05, … 1 km 08.09.2017 EMS Annual Meeting
Virtual city population & location within an agglomeration Virtual city population (Ew): Location within an agglomeration: UHI potential > 0.5 => city center UHI potential 0.2 – 0.5 => transitional area UHI potential 0 – 0.2 => outskirts UHI potential 0 => rural 08.09.2017 EMS Annual Meeting
Ideal averaging radius for UHI potential (Munich) UHI potential –> virtual population density -> UHI-effect 100m 400m 1000m 2500m 5000m Largest differences at night T-differences between Munich city & Munich airport a-e) UHI-effect removed from temperature series, f) no UHI-removal 08.09.2017 EMS Annual Meeting
Ideal averaging radius for UHI potential (Munich) UHI potential –> virtual population density -> UHI-effect 100m 400m 1000m 2500m 5000m Differences much smaller T-differences between Munich city & Munich airport a-e) UHI-effect removed from temperature series, f) no UHI-removal 08.09.2017 EMS Annual Meeting
Motivation and concept In Germany the adaptation of the impacts of climate change is framed by the German Strategy for Adaptation to Climate Change (DAS) The Federal Ministry of Transport and Digital Infrastructure (BMVI) – being responsible for the transport infra structure in Germany – funded a comprehensive national research program on safe and sustainable transport in Germany One column of this project is the “Adapting transport and infra structure to climate change and extreme weather events” 08.09.2017 EMS Annual Meeting
Evaluation Leave-one-out cross validation 1951 – 2015 1995 – 2015 [°C] BIAS MAE RMSE TAS -0.01 0.57 0.72 TMIN -0.02 0.66 0.85 TMAX 0.02 0.61 0.78 [°C] BIAS MAE RMSE TAS 0.00 0.54 0.79 TMIN -0.02 0.65 0.92 TMAX 0.02 0.60 0.85 08.09.2017 EMS Annual Meeting
Residuals & analysis 2.7.1995 TAS UHI potential 25 km² (=> Pop) Assignment for grid nodes: Continuous urban fabric: 1, discontinuous urban fabric: 0.5, green urban areas: 0.05, … UHI potential averaged over 25 grid nodes (2500 values) Required to derive virtual city population 08.09.2017 EMS Annual Meeting
Residuals & analysis 2.7.1995 TAS Results Residuals & analysis 2.7.1995 TAS Virtual city population UHI potential 1 km² (=> location) Assignment for grid nodes: Continuous urban fabric: 1, discontinuous urban fabric: 0.5, green urban areas: 0.05, … UHI potential averaged over one grid node (100 values) Required to derive the location within an agglomeration 08.09.2017 EMS Annual Meeting
Residuals & analysis 2.7.1995 TAS Results Residuals & analysis 2.7.1995 TAS Location within agglomeration City center Transitional area outskirts rural 08.09.2017 EMS Annual Meeting
Evaluation Leave-one-out cross validation (1995 – 2012) Pseudo-non-Euclidean Non-Euclidean [°C] 20 - 04 UTC 10 - 16 UTC Tmax – Tmin Bias 0.0 0.1 -0.1 MAE 0.8 0.7 1.1 RMSE 0.9 1.4 [°C] 20 - 04 UTC 10 - 16 UTC Tmax – Tmin Bias 0.1 -0.2 MAE 0.8 0.7 1.1 RMSE 1.0 0.9 1.4 10.09.2015 EMS Annual Meeting
Not accounted for UHI-effect Evaluation Leave-one-out cross validation Bias [°C] with / without UHI-effect 1995 – 2012 Bias [°C] Not accounted for UHI-effect Accounted for UHI-effect Urban -0.2 Rural 0.1 10.09.2015 EMS Annual Meeting
Evaluation Leave-one-out cross validation 1951 – 2015 Improvement due to modified distances: [°C] BIAS MAE RMSE TAS -0.01 0.57 0.72 TMIN -0.02 0.66 0.85 TMAX 0.02 0.61 0.78 Impr. RMSE [%] coast north south alps DJF JJA TAS -3 -7 -5 -11 -6 -18 -12 TMIN -2 -4 -9 -14 TMAX -10 -15 08.09.2017 EMS Annual Meeting